What Becomes Scarce After Intelligence? (manasbihani.substack.com)

In a thought-provoking essay, industry analysis reveals that the future of AI may hinge more on resources like energy, memory, and architecture rather than just advancements in model complexity. In 2023, major tech companies like Microsoft and Amazon have begun investing heavily in nuclear energy, committing around 9.8 gigawatts to power their AI operations. This movement indicates a critical shift: as AI models become increasingly open and accessible, the real value may transition towards the efficiency of hardware and energy usage, raising questions about whether the demand for intelligence will plateau. Simultaneously, the emergence of powerful open-source models, such as Moonshot’s Kimi K3, has demonstrated that significant advancements can be achieved at a fraction of the cost of traditional, closed models. These developments suggest that intelligence may soon become commoditized, shifting the focus towards efficient architectures that maximize compute power per joule—similar to how biological systems operate. As the industry grapples with the dual challenges of the energy and data ceilings, the future of AI will depend on which camp—those advocating for brute force or efficiency—can effectively navigate these constraints. Observers are urged to watch closely as the boundary between accessible intelligence and the leading-edge frontier evolves.